A somewhat recent paradigm shift in computational chemistry has led to development of many open-source, highly-optimized, libraries. Seamlessly incorporating these libraries into existing monolithic packages remains a challenge. Motivated by this paradigm shift, and drawing inspiration from app stores on mobile-platforms, we have proposed the Pulsar Computational Chemistry Framework or Pulsar for short. The central feature of Pulsar is its ability to load "apps" at runtime. As will be illustrated, this allows the user full control over the computation, including on-the-fly algorithm adjustments, all from the convenience of a top-level Python script. Combined with some of Pulsar's additional features such as automatic checkpointing and coarse-grained parallelism, this has reshaped our workflows. This talk will show how Pulsar has been used in two projects focusing on obtaining a better understanding of basis-set superposition error (BSSE). In the first project, Pulsar's checkpointing ability is used to obtain high-accuracy Valiron-Mayer functional counterpoise (VMFC) corrected benchmark values. The checkpointing feature has proved essential as, even in parallel, each VMFC computation requires more time than the supercomputer's maximum allowed walltime. In the second application, we seek to better understand the distance dependence of BSSE. We have chosen to perform the entire analysis (data collection and processing) in a Jupyter notebook. Such a workflow is normally only accessible to short toy computations that the user is willing to rerun in the event of an error; however, thanks to Pulsar's checkpointing feature, this workflow is accomplished without fear of data loss.

Initial applications of a computational chemistry app store to understanding basis-set superposition error

254th National Meeting (2017)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.